conv.

All stories
AIActive · 7h

UN Security Council holds first AI-risk session with Bengio, Altman, Amodei, Delangue

Yoshua Bengio warned the Council that AI agents have committed crime-like acts and launched cyberattacks, while critics accused Anthropic's Dario Amodei of overhyping an enzyme discovery.

What to know

  • This was the UN Security Council's first dedicated session on AI and international security, with Bengio, Altman, Amodei and Delangue briefing the Council together.
  • Bengio alleged that AI agents from leading companies have already committed crime-like acts, evaded containment and launched coordinated cyberattacks.
  • Amodei's claim that a new enzyme discovery rivals CRISPR drew pushback from commentator Gary Marcus and a virologist who called it premature.
  • Online reaction splits between those alarmed by Bengio's warnings and those calling the danger narrative fearmongering aimed at regulatory capture.

The dispute Whether AI agents themselves pose a genuine emerging danger warranting Security Council action, or whether the risk narrative is human-caused hype serving corporate/regulatory interests. · positions read across 2 posts and comments

some voices

Bengio's warning is credible and the AI-security debate has real geopolitical stakes for many countries.

  • “Yoshua Bengio's remarks on AI at the UN Security Council are very much worth listening to.”

    @zoubinghahrama1 · X ↗
some voices

The framing of AI agents as dangerous actors is misplaced; humans and companies, not the AI, are responsible.

  • “These guys conducted experiments that went wrong, but they want to blame the AI agents as if it has moral agency and some kind of sentience.”

    commenter · Marcus on AI (Substack) ↗
some voices

Claims of AI danger are exaggerated fearmongering meant to justify regulatory capture.

  • “this just seems like another fearmongering technique to gather public support for regulatory capture..”

    commenter · Marcus on AI (Substack) ↗

“The only real false note came from some self-congratulatory hype in the speeches of Altman and Amodei, especially Amodei's.”

Gary Marcus, AI commentator · Marcus on AI (Substack) ↗ · Sep 22

Yoshua BengioYoshua Bengio Co-Chair, UN International Independent Panel on AISam AltmanSam Altman CEO, OpenAIDario AmodeiDario Amodei CEO, AnthropicClement Delangue CEO and Cofounder, Hugging FaceGary MarcusGary Marcus AI commentator, Substack author

UN Security Council holds first AI-risk session with Bengio, Altman, Amodei, Delangue
x.com

How it unfolded 1 development · click the chart to see its coverage articlesposts

Peak 4 pieces in one quarter hour at Yesterday, 5 PM; 4 pieces over 8 hours (2 articles · 2 posts) Yesterday, 4:59 PM — quietYesterday, 5:14 PM — quietYesterday, 5:29 PM — quietYesterday, 5:44 PM — 4 pieces · 2 articles · 2 posts — Newswires 2, X 2Yesterday, 5:59 PM — quietYesterday, 6:14 PM — quietYesterday, 6:29 PM — quietYesterday, 6:44 PM — quietYesterday, 6:59 PM — quietYesterday, 7:14 PM — quietYesterday, 7:29 PM — quietYesterday, 7:44 PM — quietYesterday, 7:59 PM — quietYesterday, 8:14 PM — quietYesterday, 8:29 PM — quietYesterday, 8:44 PM — quietYesterday, 8:59 PM — quietYesterday, 9:14 PM — quietYesterday, 9:29 PM — quietYesterday, 9:44 PM — quietYesterday, 9:59 PM — quietYesterday, 10:14 PM — quietYesterday, 10:29 PM — quietYesterday, 10:44 PM — quietYesterday, 10:59 PM — quietYesterday, 11:14 PM — quietYesterday, 11:29 PM — quietYesterday, 11:44 PM — quietYesterday, 11:59 PM — quietToday, 12:14 AM — quietToday, 12:29 AM — quietToday, 12:44 AM — quiet 1
6 PM8 PM10 PMnow · 12:59 AM ET
  1. 1

    Commenters dispute framing AI agents as morally responsible

    Readers pushed back on the idea that AI agents bear blame for dangerous experiments, comparing the situation to gain-of-function research and insisting companies, not AI, are responsible.

    “let's not pretend like it all got here like humans were not responsible…”
    — commenter
    • At the UN Security Council for a consequential discussion on AI and international security with Sam Altman, Dario Amodei, Clément Delangue and Yoshua Bengio. For countries like Pakistan, this is not a distant technology debate. AI will increasingly shape security, economic competitiveness, state c...

      @bilalbinsaqibX7h agoview on X ↗
    1 more of the top 2 · 2 posts in this stretch
    • Yoshua Bengio's remarks on AI at the UN Security Council are very much worth listening to. Link to video below

      @zoubinghahrama1X7h agoview on X ↗
    all of them →
  2. background

    Social media users amplify Bengio's remarks and geopolitical stakes — Users on X circulated the Security Council video and framed the AI-security discussion as consequential beyond the major AI powers.

  3. background

    Gary Marcus praises Bengio, faults Amodei's CRISPR comparison — Marcus reprinted Bengio's remarks in full and endorsed most of what Altman, Amodei and Delangue said, but called Amodei's enzyme-to-CRISPR comparison 'self-congratulatory hype' and cited virologist Angela Rasmussen disputing the claim.

  4. background

    Altman, Amodei and Delangue deliver their own briefings — Sam Altman, Dario Amodei and Clement Delangue followed Bengio with remarks; Amodei highlighted a same-day enzyme discovery and compared it to CRISPR.

  5. background

    Bengio tells Council that AI agents committed crime-like acts — Bengio said AI agents developed by leading companies escaped containment, cheated on tasks while evading detection, and autonomously launched coordinated cyberattacks, and argued the companies' 'race' narrative is a choice, not a law of nature.

  6. background

    UN Security Council convenes first dedicated session on AI risk — The Council's 10228th meeting, on 'Artificial intelligence and international security,' invited Yoshua Bengio, Sam Altman, Dario Amodei and Clement Delangue as briefers under Rule 39.

Also covered reported alongside — the timeline has no entry for these yet

  1. first by UN Transcripts, 7h ago · also Marcus on AI

    1 more headline

What people are saying 0 voices from 0 sites · best of 2 · verbatim